Fail Friday – Stories of when AI goes wrong: Kite
Fail Friday – Every Friday we share stories and examples of AI fails that we can all learn from.
The case of Kite
When researching for stories and data for the last post we came across the detailed experience of Kite. Kite was a start-up that used AI to help developers write code with an ambitious goal of improving productivity for developers by 10x. Their story is here and worth a read for the reminders and lessons of the difficulty of starting a business and how chasing novel AI can send teams down the wrong path: https://www.kite.com/blog/product/kite-is-saying-farewell/
There is self-reflection in their story but we agree with the cause of their failure:
“We failed to build a business because our product did not monetize, and it took too long to figure that out.”
We agree because of the last part of this quote …It took to long for them to figure it out. They shared what they did wrong and this is the lesson for all of us building AI for others:
“First we built our team, then the product, then distribution, and then monetization”
Here’s our argument. AI is about the people that will use the AI-powered tools and features, it is NOT about the developers of the tools. This is what we call “People-first AI.” While People-first AI appears straight forward, it is difficult for development teams and companies to have the discipline in the AI race to go from an idea to building a product their users will love. It’s common for creators to fall in love with their hypothesis, recruit people, raise investment, and spend that money before ever deeply testing this hypothesis. Teams can go deep and far before realizing they went down the wrong path and burned all of their cash.
People-first AI: Obsess about the people that will use your future AI product. How can AI amplify peoples’ skills and improve the productivity of their tasks so much that it becomes Obvious to them they MUST use this AI? Interview, include, and have people use the AI until they can’t stop using it.
This is the failure of Kite. Over 7 years Kite spent millions on building a team, building technology, building a sales channel, and recruiting 500,000 monthly-active developers first…then they spent time with their users to figure out what they would pay to use. The people that would user their product and their buyers had to come first. Unfortunately the Kite team learned this too late and failed. Also, the team also believed the 10x productivity gain was necessary. Today there are several tools and several research papers that share that there is value in AI-powered coding assistants without hitting this number. Maybe the 10x number should have been a goal, like going to mars. Instead of spending all of their resources aiming for mars - aim for space first, next the moon, and then set your sights on the ultimate goal.
